Linking Decisions and Information for

Organizational Performance

A Client Study by Thomas H. Davenport

October 2008

Linking Decisions and Information for Organizational Performance Thomas H. Davenport

We are entering a new period in the era of business information systems. For most of this five-decade period (beginning in the mid-1950s), the primary focus has been on automating core business processes. The era began with custom-developed narrow-purpose applications and concluded with broad enterprise system packages provided by external vendors, but the purpose was the same: develop control and efficiency over processes by automating and capturing information from key business transactions. Whether a general ledger entry, a customer order captured, r a vacation balance debited, the transaction has been the primary unit around owhich this world revolved. By now, however, we have largely won the transaction war, and a new front is necessary in the war to manage information. While mop-up and maintenance activity on the transaction and application fronts are still necessary, most large rganizations have the basic application functionality they need, and their otransactions have largely been automated. What remains to be done is simply to fulfill the promise of the information age from its beginning. That is to use the accumulated information from transaction systems to optimize the management of the business. From the beginning, the idea was that better information would lead to better decisions and better ways of managing organizational processes. Whether this idea was called decision support, executive support, online analytical processing, or business intelligence, there was always another goal waiting to be achieved. Because organizations’ efforts and attentions ere being spent on automation, to improve decision-making was never the wprimary focus. Today it has taken center stage. If transactions were the focus of the automation era, then decisions must be the focus of this optimization era. If the goal of better information and better analysis is ultimately better decisions and actions taken based on them, we must renew our focus on decisions and their linkage to information. Businesses need to address how decisions are made and executed, how they can be improved, and how information is used to support them. And they must look at all types of decisions: from strategic lanning decisions made by senior management to every day operational decisions pmade by employees on the front line, or automated by back end systems. There are obvious benefits to improving decision processes. Many organizations suffer from poor decision processes and outcomes. There is a growing body of knowledge on optimal decision processes and decision biases to avoid, but it is rarely applied within organizations. Information that is available to inform decisions isn’t used, or information is captured and managed that is unsuited for decision purposes. Decisions frequently take too long to make, and involve either too many or too few individuals. We hardly know the extent of the problem and the potential

benefits, for most organizations don’t identify, assign clear responsibility for, or track the results of their key decisions. This reporti describes a study of attempts by organizations to improve decision-making through the use of information, among other interventions. I spoke with 31 managers in 26 organizations about specific initiatives their organizations had undertaken to improve decisions or decision processes. In each interview I asked about why the initiative had been undertaken, how the decision process varied efore and after the intervention, and what steps were taken to provide the decision rocess and decision-makers with better or more trusted information. bp Focusing on Information and Decisions My intent was to understand how information is being applied to improve decision-making in a broad range of contexts. A list of the decision types and organizational contexts is provided in Exhibit 1. Most of the decisions listed are made frequently and involve core business processes of the organization. I sought out such core processes because it seemed that they would be the most likely to be the subject of initiatives to supply information for decisions. The organizations interviewed were dentified through secondary literature and discussions with IBM and other nformation technology consultants an s. ii d vendorExhibit 1 Types of Decisions Studied 1. Supply chain and financial decisions in electronics distributor n 2. Credit and risk decisions in a money center bank rant chai3. Marketing and performance management decisions in a fast food restau4. Performance management and supply chain decisions in a vehicle manufacturer t store chain ent organization 5. Merchandising and loyalty decisions in a retail departmen6. New product development decisions in an educational assessm7. Credit and risk decisions in a consumer finance company e company ompany 8. Energy project credit decisions in an energy financ9. Real estate finance decisions in a commercial real estate financing cirm 10. Sales decisions in an IT product and service firm surance fer 11. Retail financial services decisions in a banking and in12. Claims and disease management decisions in a health insuror trict 13. Project estimation decisions in a defense contract14. Student performance decisions in an urban school disuipment firm hospital chain 15. Pricing decisions in an industrial eq16. Physician drug ordering decisions in a17. Critical care decisions in a hospital 18. Logistical decisions in a trucking firm insurer 19. Pricing decisions in a carpeting manufacturer 20. Financial and disease management decisions in a health21. Organ donation decisions in an organ sharing network

22. Student performance decisions in a public university m23. Small business insurance underwriting and delivery in a ajor insurance firm 24. Oil drilling decisions in a mid-size integrated oil company 5. New greeting card decisions at a greeting card company 226. Automobile financing decisions in a sales and financing company While most of the managers interviewed were comfortable with talking about attempts to bring about better decisions, the topic was not yet “top of mind” in most companies. It was clear in the discussions that most firms had not focused consciously on better decisions as an area for business improvement. Some had not initially viewed their efforts as decision-oriented; this was true, for example, at Educational Testing Service, which was attempting to improve its new product evelopment processes. The manager interviewed stated, however, that the key dissue in the process was making decisions about which products to develop. There were some exceptions, however, to the “invisibility” of decisions. Citigroup and another global retail bank, for example, had created “decision management” groups that focus particularly on analytical and quantitative decision processes. Procter & Gamble has renamed its IT organization “Information and Decision Solutions,” and the organization contains substantial numbers of analysts who assist decision-makers with analyses and fact-based decision processes. While these organizations are moving toward a stronger focus on decision-making, most do not seem to have broad agendas in place for connecting information and decisions in eneral, though they may have particular decision emphases such as greater use of nalytics or automated decisions. ga Linking Decisions and Information How do organizations ensure that decisions are made on the basis of the best possible information, and that the right information is gathered and analyzed to support decision processes? There are at least three different levels of relationship etween information and decision-making, each of which were present in the rganizations interviewed for this study (Ebo xhibit 2). Loosely-Coupled Information and Decisions—Perhaps the most common approach to linking information and decision-making is to loosely couple the two. That is, organizations often make information broadly accessible to analysts and decision-makers for application to decisions, along with tools to manipulate and display the information. The information usually involves a particular business domain—finance, marketing, sales, or overall performance management, for example. However, it is intended to inform a range of possible decisions, and the actual use of the information for any particular decision is voluntary and based on individual nitiative. There is no monitoring of what information is used for what decisions, ither before or after decisions are taken. ie

This loosely-coupled approach would characterize most organizations’ approaches to business intelligence, or what was previously called “decision support.ii” Data suitable for analysis and decision-making is extracted from transaction systems, and made available in a data warehouse or mart. Standard reports are produced, perhaps in easier-to-understand “scorecard” or “dashboard” formats. The appeal of this approach is that providers of information can supply it without regard to difficult and sensitive issues such as managerial psychology, organizational politics, and decision rights. In such decision environments, more structure or automation may not be appropriate or necessary. In this model it is not the task of the information provider (nor anyone else) to ensure that the decision is informed by the information or made well. Also appealing is that an information infrastructure an support a variety of decisions, which is productive and efficient for information roviders. cp Exhibit 2 Three Approaches to Linking Information and Decisions However, while it does not directly address managerial decision processes, this loosely-coupled approach still presents many informational challenges. In order to provide information suitable for decision-making, information must usually be integrated from multiple source systems and of high quality. Organizations also struggle with developing a “single version of the truth,” so that information for ecisions is consistent across the organization. It is easy for multiple versions of eports and data entities to proliferate across large, complex organizations.

More Decisions

Loosely-Coupled Information Decision

Structured Human

Information Decision

Automated

Information Decision

Requires linkage to decision process &

behaviors

Requires tight process/system integration

Requires information

infrastructure

+

+

dr

I found several examples of this loosely-coupled approach in the study, and others are common throughout the business intelligence literature. For example, BlueCross BlueShield of Tennessee created a “Financial Data Mart” to support a variety of financial decisions. The data mart was supplied with high-quality data on operations, product utilization trends, and financial results across the organization. Considerable efforts were expended to ensure “one version of the truth.” Training was provided on how to use the system and how to access commonly-used reports and data cubes. The IT organization is of the impression that the primary users were much more able to create and apply reports that informed their decisions. Of course, since the specific decisions to be made from the data mart were not directly linked, the value to improvements in decision-making remains an intangible benefit to the organization. Other examples of this sort of decision/information relationship in the study included a student performance analysis and reporting system in the Miami school district, and a somewhat broader business intelligence system at at a large Australian university. Both were focused on getting a better understanding of student performance; the university’s system also addressed research grants, financial management, and human resources. Both were viewed as successful by the anagers interviewed, and they required similar types of effort and investment as mthat of BlueCross BlueShield of Tennessee. Of course, making these loosely-coupled decision environments work requires much more than simply making information available. Firms that had successfully improved decision-making described such approaches as a strong alignment between IT organizations and business units, approaches to developing the apabilities of users, and clarity on the business objectives of data warehouses and arts. cm Structured Human Decision Environments—Some organizations I interviewed had a narrower focus on particular decisions, but tried to create an overall decision-making environment that went beyond just establishing an information infrastructure. In this approach, the decision at issue is still made entirely by human managers or professionals, but specific efforts have been made to improve targeted ecision processes or contexts by determining the specific information and other dprocess resources needed to make better decisions faster. The advantage of this approach is that these additional efforts created a stronger linkage between the information and the relevant decisions, making it more likely to be used effectively. The challenges of this approach relative to the loosely-coupled one are its narrower focus on particular decisions, and the additional effort needed o create the decision environment. If the decision is an important one for torganizational success, however, it may be worth the additional effort. The type of additional support for decision-making varied widely across different examples in the study. In some cases analytical tools and capabilities provided the additional decision support, as in the case of pricing decisions at Parker Hannefin,

an industrial equipment firm; marketing and performance management decisions in CKE Restaurants, a fast food restaurant chain; and merchandising and loyalty decisions at Dillard’s, a retail department store chain. In Parker Hannefin’s pricing example, salespeople were provided with analysis yielding a target price, a floor price, and a ceiling price based on analysis of previous sales and segmentation of the relative differentiation of the product being sold. At CKE Restaurants, randomized testing analyses were used to support new product marketing decisions, and econometric models provided explanations of which factors drove changes in weekly sales results. At Dillard’s, predictive models of sales for particular brands were used to order merchandise for stores and regions. The use of such analytical pproaches is an increasingly common feature of information-rich business aenvironments.iii However, in all three examples, analytics were not the only approach to improving the decision environment; there were also investments in establishing accurate, trusted information, along with a focus on organizational and behavioral techniques being employed. At Parker Hannefin, the company also felt the need to create new divisional pricing manager roles to ensure that salespeople understood the new pricing approaches and adopted them successfully. At CKE Restaurants, the Chief Information Officer employed principles from cognitive science research to maximize the likelihood that executives would notice and understand key information. At Dillard’s, information providers worked closely with early adopters of the new decision approach to identify ways to spread the use of the approach to less analytically-oriented merchandisers. And in all three cases, an information nfrastructure was put in place with a particular focus on delivering the right iinformation needed to support improved decision making. In other cases, organizations employed tools to provide additional structure around the decision process. At GE Energy Finance, an analyst interviewed senior executives to understand the factors they used in making financing decisions, and developed a model of the factors they employed using conjoint analysis (an analytical technique usually employed to understand customer preferences in marketing). While senior executives still actually make the decisions, the model has been very helpful to less experienced employees in preparing their financing proposals, and research has shown that decisions made using the factors uncovered n the analysis are substantially more successful than those made with unaided iexperience and intuition. At Daimler Truck, decisions around performance management, supply chain and other operational issues were incorporated into a broader context. The company had adopted the A3 problem-solving approach as used successfully by Toyotaiv. The approach structures a set of problem resolution and action steps on two sheets of paper (A3 size in Japan). The approach ensures that information and decisions result in improved business performance. Hallmark used considerable market research-based information, and a decision-structuring framework called the Value Star, to assess whether a new line of lower-cost greeting cards provided sufficient

value to customers. The Value Star represents customer value as a combination of five factors: Equity, Experience, Energy, Product and Money. Technology also provides support for these structured human decisions. Scorecards and specialized information displays provide just the information needed by decision-makers, and no other. Recommendation systems based on algorithms or ules provide a recommended decision, but in these types of decisions they can rusually be overridden by human decision-makers. Given the breadth of the actions and tools that organizations adopt to better connect information and decisions and the challenges of addressing managerial behavior, this approach can’t be adopted for all decisions. The decisions selected for this sort of intervention must be particularly critical to organizational success. That is, they hould involve strategic issues or important everyday decisions that drive business erformance. sp Automated Decisions—The closest linkages between information and decisions usually come when decisions are made by computer.v When it is critical for information to be applied to a decision in a structured, formulaic fashion, the answer is automated decision systems. While the majority of press and visibility came to artificial intelligence and expert systems two decades ago, many firms have quietly implemented more straightforward automated decision-making in a variety of business domains. In order to optimize operational decision-making, companies have embedded decision rules and algorithms into key business processes. In doing so, many have achieved greater speed, decision accuracy, and better customer service. While human experts, of course, design the system in the first place, with utomated decisions they are not the primary decision-makers, and usually come ainto play only in handling exceptions. Automated decision-making systems are not a new idea—they first took hold, for example, in “yield management” systems in airlines that made automated pricing decisions in the early 1980s—but the applications for the idea are expanding significantly. After yield management, automated decision-making then became pervasive in the financial services industry, and is still most common there. In investment banking, these systems are behind the rise of program trading of equities, currencies, and other financial assets. For most consumers, the primary impact of automated decision-making is in the realm of credit approval. Credit scores are used to extend or deny credit to individuals applying for mortgages, credit cards, and other forms of debt. Though credit scoring has been criticized for being overly simplistic, it has certainly made the process more rapid and efficient, nd there is no longer any doubt that credit score information is being applied to adecisions that it can inform. In this study, the automated decision activities were at Citigroup, another global retail bank, a large, privately held automobile sales and financing firm, and Zurich Insurance North America. All four organizations had institutionalized the process of

developing and using automated decision systems. Zurich Insurance had begun using the approach on individual-level underwriting decisions, but had extended it to more complex small business policies. The company also built a special portal for its agents to use in entering data from the system and receiving results. Citigroup and the other bank with an automated decision focus are primarily focused on automated credit and lending decisions. A large automotive leasing and financing firm is reengineering several of its business processes for automotive financing sing automated decisions to the improve efficiency and effectiveness of recurring ufinancing decisions. Again, in all four cases, there were significant investments made in the underlying information infrastructure. As decisions become automated, it becomes ncreasingly important to ensure that the information used is complete and accurate isince there is no human involved for fact checking. Of course, the development of automated decision systems is time-consuming and expensive. Firms must be very selective in deciding which decisions to automate. The decision process must be sufficiently structured and reducible to rules or algorithms, and a complete and direct linkage to all of the information needed must be created. There must also be frequent review of the decision rules or algorithms to ensure that they continue to produce the right decision outcomes. The automotive leasing and financing firm has a clear set of criteria to identify the processes that are most likely to benefit from automated decisions, and is committed to reviewing them frequently for needed revisions. The firm also integrates multiple information technologies to support the reengineered processes, including a workflow system for coordinating process flow, and a rules engine to store and execute business rules. Despite these challenges, automated decision-aking provides the closest possible link between decisions and information, and e can expect such systems to continue to grow in popularity and effectiveness. mw A Process for Connecting Decisions and Information Given these three options for relating decisions to the information that informs them, organizations can follow a process for establishing and maintaining the onnection. The process may vary somewhat with the particular ecision/information linkage that the ocd rganization follows. Step 1: Strategic Focus on Key Decisions—Since connecting information and decision-making often requires a major investment of resources, it’s important to ensure that any decision selected for intervention is actually important to the organization’s strategy and performance. Therefore, a reasonable first step is for an organization’s executives to discuss the strategy and determine what decisions are important to its successful execution. It may not be necessary to select decisions that are the most important of all, but no organization should waste time and energy on decisions that don’t matter. And at least in retrospect, the choice of decisions for intervention often seemed obvious in the examples surveyed.

For example, Irish Life, a financial services company with major business units in life insurance and banking, concluded that it needed to become closer to its customers and to offer a more integrated range of financial services to them. Its management team decided that decisions around which products to offer which customers were critical to its strategy. After identifying the decision, the company embarked upon a series of efforts to pull together the information environment that would make an integrated view of customers possible. Better decision-making by ustomers was also a goal, in that the new online environment would make it possible cfor them to see all of their holdings in one place. Partners HealthCare System, an academic medical center in Boston, discovered in the 1990s that it had unacceptably high levels of medical error. The organization’s leaders decided that a key decision process was that in which physicians decided which drugs, tests, treatments, and referrals to administer to patients. This process, and information systems that address it, is known in the health care industry as “physician order entry.”vi The importance of the decision to Partners’ primary ission of better patient care is illustrated by the successful result of the order mentry intervention: a 55% reduction in “adverse drug events.” Organizations that do not address this strategic step first in their attempts to provide information for decision-making face a key risk. They may end up building information environments that don’t help decision processes in business-critical areas. They may not be able to determine whether their efforts were worth the investment of money and time. Still, there are many organizations, including some n this study, that had embarked upon substantial information provision projects ithout any strategic clarity iw about what particular decisions they support. Step 2: Information Provision—Given an important decision that’s key to an organization’s strategy, organizations must begin to provide information for it. In loosely-coupled relationships between decisions and information, this is appropriately the second step in the process; if the information and the decision are more closely coupled (in structured human decision processes or automated decisions), it may be more appropriate to first undertake Step 3 below involving decision design. The order in which information provision and decision design take place also varies by the amount of time it’s estimated to take to make information available to the decision. The provision of information may lead to the development of a data warehouse, a more focused data mart, or a specific analytical application. ither way, the accuracy and completeness of information will have a direct impact Eon the ease and effectiveness of the following steps. The inf series of natural questioormation provision step best begins by simply asking a

• ision? ns about the decision, such as: What information is required to support the dec• How accurate does the information need to be?

• d supplying the What’s the most efficient process for collecting, generating, aninformation? • In what timeframe does the information need to be supplied? For example, United Health Care, a large health insurer, concluded that its most important decisions were in the areas of claims and such specific activities as claims adjudication, disease management, and claims payment. In order to address these decisions, United concluded that it needed to take a bottom-up look at claims information—how the information is gathered and stored around the company. It is constructing a large enterprise warehouse of claims information, and is also developing what it calls “data communities” ”— a set of inter-related data marts that share a common business needs such as claims across multiple business units. hese data communities can be connected or used stand alone to solve specific Tbusiness problems related to claims, providers, enrollment, conditions, etc. The challenge of the information provision step—particularly if it is undertaken before decision design—is to keep in mind the specific decisions the information is to inform. It is all too easy to become wrapped up in information management ssues and to lose sight of the decisions involved. Organizations need to make sure hey have an informatiit on agenda that is being driven by their business objectives.

Step 3: Decision Design—In this step, the key aspects of the context for the decision being made are designed, or at least evolve in a preferred direction. Important considerations in the design process include identification of the roles different individuals will play in the decision, the level of structure for the decision, the ability f human decision-makers to process the relevant information, and the roles of ohumans vs. computers in the decision process. In the study of 26 decisions, I found a few organizations in which decisions were explicitly designed. GE Energy Finance, where the factors driving executive decisions were explicitly modeled and communicated to decision-makers, is one example of a consciously-designed decision process. Partners HealthCare’s physician order entry system is another. An automotive leasing and financing firm is redesigning many automotive financing processes, and is addressing the key decisions made in those processes at the same time. More frequently, however, the decision context had simply evolved over time with multiple interventions. For example, at a mid-size oil company, the decision involving in which areas to drill for new oil had been the subject of several incremental improvements over time intended to bring greater structure and effectiveness to the decision. The company had invested in a formal “Prospect Evaluation Sheet” that recorded the story and history of how the lead progressed to its current prospect level. The company had also depicted the exploration decision-making process in a visual format, which greatly enhanced the ability of participants to understand their roles, responsibilities, and interactions throughout the process. Still, despite the company’s efforts to better structure the decision and a massive

amount of seismic and geological information, the decision process remained more iterative and subjective than some managers would have preferred. In automated decision processes, organizations must explicitly design not only the rules that will be embedded in the automated decision system, but also the performance objectives for the process and the role for human experts in designing and operating the system. In Zurich North America’s insurance underwriting decision process, the company designed the new process to optimize the cost, time, and quality and consistency of policy underwriting, as well as measures of how long it takes to add or change a rule and modify the underwriting criteria. They followed a rule of thumb for utilizing underwriters to keep the best performers away from routine underwriting. Underwriters should instead do “portfolio management”—looking across all the rules, monitoring performance and looking for new business areas. They also specified the conditions under which human underwriters would ecome involved in handling exceptions, e.g., those involving high dollar amounts or issing data. bm Step 4: Decision Execution—The final step in connecting information and decisions is to operate and manage the decision process over time, and to ensure that decision-makers use information to make better decisions. This almost certainly involves training of users on the available data, on the use of systems to access the data, and perhaps on the factors to consider in decision-making. BlueCross BlueShield of Tennessee, for example, spent considerable resources designing a training program for financially-focused users of the business intelligence system, and then edesigned the training later to address changes in the business and the available rdata. Those who are responsible for ensuring effective use of the information in decision processes may also want to enlist influential executives as users. In the Miami school district, the frequent and aggressive use of the system by the superintendent led principals and teachers to make more use of it as well. At the Australian niversity, the school that used the business intelligence system most effectively uhad an influential user in the dean of the school. Organizations will also need to modify and improve their decision processes and systems over time. At Partners HealthCare, the physicians can override system recommendations that they do not agree with. The institution monitors which treatment decisions are frequently overridden to determine whether they are faulty or unnecessary. Partners also employs an online discussion system to allow expert hysicians to discuss and decide upon new treatments to be added to the order ntry system over time. pe

A Stage Model for Decision Management It is possible to place the different approaches of firms to connecting information and decision-making into a stage model (Exhibit 3). The five stages differ in terms of he specific focus on decisions and how oriented the organization is toward linking tinformation to decision-making. In Stage 1, firms are still wrestling with providing basic transaction information, and have not really focused on decisions at all. The information addresses all decisions equally, i.e., not at all. While some of the firms I interviewed for this study were still focused on putting transaction information in place, they all had some ambition to provide that information for decision processes. Therefore, I would not classify hem as Stage 1 organizations. In the world at large, however, Stage 1 organizations tare quite common. In Stage 2, organizations lay the foundation for information relating to decision-making. They may build data warehouses and marts, for example, to support a variety of decisions, though their focus is somewhat narrower than Stage 1 in terms of the breadth of decisions. As described above, they may be trying to facilitate financial decisions or customer-oriented decisions, but not all types. Most large firms are at the Stage 2 level. Stage 1 and Stage 2 organizations are primarily focused on the supply of information. However, just as with automating transaction processes, there can be different degrees of what has been accomplished at Stage 2. Companies that have made a greater investment in the accuracy and completeness f information will find it easier to reach the later stages, and will drive better oresults from those investments. At Stage 3, firms are just beginning to focus on decision-making, and consequently they tend to address one decision, or a small set of decision processes. As noted above, some may think of their efforts not as improving decisions, but rather as addressing a key business process. Educational Testing Service, for example, addressed new product development decisions as a business process improvement project, and the mid-sized oil company addressed oil drilling decisions in the same rocess improvement context. Companies can sometimes enter Stage 3 without psignificant investment in Stage 2 because of the narrower focus. In Stage 4, organizations begin to address a broader set of decisions by focusing on the decision context. They may have an overall perspective related to human cognitive processes (as in CKE Restaurants), to applying information and decisions in the context of a problem-solving approach (as in Daimler Truck), or a broad nalytical orientation (as in firms such as Harrah’s Entertainment and Capital One athat “compete on analytics,” but were not included in this study.) Stage 5 organizations have determined that effective decision-making is critical to their success, and they have created organizations and processes for improving it on a substantial scale. The two financial services firms, Citigroup and another retail

bank, that had created “Decision Management” groups fall into this category. Even at Stage 5, however, the firms had not yet prioritized all decisions for intervention across the organization, and considerable latitude for decision improvement remained. Stage 4 and 5 organizations are still interested in the supply of nformation, but they also address the demand for information in the decision iprocess, and focus on the broader context in which it is applied. aving accurate, trusted information is critical to reaching Stage 4 or 5, and acts as he foundation for driving effective results and optimizing business decisions. Ht Exhibit 3 A Stage Model of Decision Management

Looking Ahead in Decision Management We are only in the very early stages of improving decision-making and making better use of information in decision processes. As organizations move in this direction, we will undoubtedly learn about new approaches to linking information and decisions, and to improving the broader context for decision-making. We will also probably see new information technologies that attempt to structure and improve decision processes. While we now have many of the technological components for better decisions—including data warehouses, business intelligence tools, workflow systems, decision rule engines, and so forth—these components are not yet well-integrated, and organizations are unsure about how they fit together. erhaps in the future we will have “decision management” systems that incorporate Pall these capabilities as well as others. The primary obstacle to decision improvement efforts will be traditional understandings of management responsibility for decision-making. If we view

decisions as an individual managerial prerogative—not subject to review or improvement—then we can make little progress in making them better. Many firms ave implicitly treated decision-making in this fashion, and hence they will have hdifficulty with interventions intended to improve decision-making performance. Decision-making has always been viewed as one of the most important activities of everyone in an organization, from executives and managers to front and back office employees handling everyday customer interactions and transactions, and it is difficult to overestimate the value of improving it. Decisions have impact on every aspect of organizational performance, in both strategic and tactical domains. Organizations have too much at stake to continue with the poor decision processes f the past. It is time for them to address better decision-making as one of the last—nd most important—frontiers of business performance improvement. oa i The research was sponsored by IBM’s Information on Demand business unit. Thanks to Marc Andrews, Dena deBin, Holly Haswell, and other IBM contacts for their guidance and suggestions. ii Decision support systems were first popularly described in the early 1970s. See Anthony Gorry and Michael S. Scott-Morton, “A Framework for Information Systems”, Sloan Management Review, 13, 1, Fall 1971, 56-79. An online history of d Power’s DSS Resources site: ecision support systems is available at Danhttp://dssresources.com/history/dsshistory.html iii Thomas H. Davenport and Jeanne Harris, Competing on Analytics: The New Science o also Ian Ayres, Supercrunchers: f Winning, Harvard Business School Press, 2007;Why Thinking By Numbers is the New Way to Be Smart, Bantam, 2007. iv ight Things Done: A itute, 2006. The A3 approach is described in Pascal Dennis, Getting the RLeader's Guide to Planning and Execution, Lean Enterprise Instv See, for example, Thomas H. Davenport and Jeanne G. Harris, “Automated Decision-Making Comes of Age,” with Jeanne Harris, MIT Sloan Management Review, Summer 2

. 005, 46:4, 83-89; also James Taylor and Neil Raden, Smart Enough Systems: How to

Deliver Competitive Advantage by Automating Hidden Decisions, Prentice-Hall, 2007vi For more on the Partners example, see Thomas H. Davenport and John Glaser, “Just-In-Time Delivery Comes to Knowledge Management,” Harvard Business Review, July 2002.